Safe and efficient learning

Production systems must include tests checking that operating and training conditions match. Models need to be hardened against spurious inputs, high or wrong uncertainty estimates and deliberate adversarial attacks. Due to acquisistion or labelling costs, oftentimes this must be achieved with little data.

Content series in this area

Efficient RL

Recent and popular advances in Reinforcement Learning are known to be data-hungry. Attempts to handle this deficit include developing complex simulators while improving the sim2real transition of models, explicitly modelling the …

Robustness in ML

The robustness of a neural network classifier is a measure of the stability of its predicted classes against input perturbations. We here investigate and review recent developments in the field.

Safety in learned controllers

Safety constraints for learned controllers with known dynamics.

Data efficiency

Numerous applications of Machine Learning in business intelligence, process optimization, product development or sales do not benefit from true “big data”. Blind application of state-of-the-art methods can lead to catastrophic …

Data valuation

Attributions of value to training samples can be used to examine data, improve data acquisition, debug and improve models or compensate providers. Recent developments in the field enable principled definitions of “fair” value …

Check all of our work